Commodity Clouds Vs. Differentiated Clouds

Randy Bias (@randybias, CEO of Cloudscaling) tweeted earlier today that -- compared to things like a Vblock -- the clouds that he was capable of building were about 1/10th the cost.

Now, that's an interesting discussion, to be sure. 

And I'm not here to berate Randy -- or anyone else -- who agrees with this line of thinking.  Randy appears to be a smart guy, and very experienced as well.

It's not a new thought -- the broader industry debate around commodity vs. differentiated has been around for a very long time indeed, and isn't going away anytime soon.  It's moved from hardware, to software and now to the cloud.

But how does this historic debate play here?  And especially in the context of service providers who are looking at this more as a business proposition, as opposed to an intellectual debate?

A Few Disclaimers

We really don't have any head-to-head standardized comparisons to go look at, but I will theoretically grant a key point: raw computing ingredients (compute, memory, physical storage, etc.) are damn cheap when you go shopping in the bargain section of the IT supermarket.

Heck, the last terabyte of storage I bought for my home was less than $90.  I bet it's even cheaper now.  As a matter of fact, it was an Iomega product (an EMC division), so we're no strangers to the allure of cheap and cheerful tech.

But there's a bit more to the picture that is worth considering.

Someone Has To Assemble The Pieces

Randy's firm is what I call a "cloud builder".  They understand the requirements, select the components, assemble the pieces, build some operational tools, and eventually hand it over to the service provider.

I am presuming that Randy's services (or anyone else like him) are not free.  Given his level of expertise, I would expect that he deserves to command a premium in the market.  So I think a more useful comparison is probably 'cloud delivered ready for use' rather than 'prices for raw ingredients'. 

Make no mistake -- selecting components, integrating them, qualifying their boundaries, characterizing their behavior -- that's a lot of heavy lifting.  When you buy something pre-integrated like a Vblock, all that effort is already baked in.

Who Ya Gonna Call?

Let's say something breaks.  I'm not talking about a simple component failure, I'm talking about several pieces that aren't interacting the way they're supposed to.  None of the componentry vendors have invested in support models that can handle how things interact, rather than the individual pieces themselves.

Add in the inevitable upgrades, patches, enhancements, etc. -- more heavy lifting.

If the end result is to be used for commercial purposes, those are costs that have to be identified and figured in somewhere.  With the Vblock model, those costs are certain and visible -- less so with other commodity-based models.

So we've gone from "cost of ingredients" to "cost of delivered cloud" to "cost to support and operate". 

There's more, but I think you get the picture ...

What Can It Do?

Take something like a Vblock.  You'll see a lot of differentiated technology.  Differentiated virtualization from VMware.  Differentiated converged server and storage from Cisco.  Differentiated storage, backup, replication, management and security from EMC.

All of that differentiation was done with a purpose: more performance, better availability, tighter integration, stringent security, etc.  As a result, it can support mainstream IT workloads using a cloud operational model -- just about all of them.  That's useful, as we'll see in a moment.

If we take even one or two aspects -- like "performance" or "availability" -- unfortunately there are no good metrics for independent comparisons.  BTW, these terms have subtly different meanings when discussing workloads specific to enterprise IT vs. generic compute images.

Speaking strictly on behalf of EMC, we are quite familiar with what commodity clouds can do, and have built a few of our own.  We have a good idea as to what they do well, and what they don't do well.

In all fairness, we do encounter use cases (and associated business models) where -- yes -- the commodity-based cloud is an attractive option.  But -- at present -- there's not a lot of these use cases to go around, and there doesn't appear that there's a lot of money to be made, since -- well -- they're commodity clouds, and hence undifferentiated.

Which brings me to ...

Who's Going To Want It?

By far, the largest target market for cloud services are traditional enterprise IT organizations. 

And I am continually amazed at how so many people think that "it's cheaper!" will somehow overcome all the other valid concerns these people have when considering any form of an external IT service.

Not to make an unpleasant analogy, but consider health care -- especially in more developed countries.  Here in the USA, we're very concerned with controlling rising health care costs -- but not at the price of substandard care!

Are there enterprise-class workloads that could possible run in a commodity cloud?  Perhaps, but examples are hard to come by.  Certainly not the meat-and-potatoes stuff like large databases and critical business processes.  At least, not in the near term :-)

The Service Provider Business Angle

Given my exposure to SP business models, I can make a strong case for many of them to have a rock-bottom, low-cost IaaS offering.  If someone is shopping price -- and price alone -- you need to have something to get into the discussion.

As a standalone business model though, I find it singularly unattractive.  There are a limited number of paying customers, more and more people are entering the market, and prices continue to fall.

However, as an "on ramp" to more differentiated services -- and hence more profitable -- it is a far more attractive proposition.  Create an environment where it's easy to move people up to more -- more performance, more availability, more security, more control, more functionality, etc. -- and do so seamlessly and non-disruptively.

The real question in my mind?  Does it make sense to do this with one infrastructure and model that can flex upwards, or invest in two distinct architectural and management models -- one optimized for sheer cost, and one optimized for the delivery of richer services?

My argument has been the former, and -- so far -- it's getting a fair amount of acceptance.

That being said, it's still early days :-)

With All Due Respect

I think the commodity vs. differentiated discussion is a good one.  It was good 20 years ago, and it'll probably be good 20 years from now.

The "good enough" school keeps getting better.  What used to require differentiated technology can sometimes be done using a commodity approach.  However, most experienced practitioners have learned to look beyond the simple prices of the ingredients.

That being said, differentiation hasn't stopped either -- it keeps getting better as well. 

The real challenge for service providers?  Keeping ahead of both waves :-)

Comments

[josephmartins]

I quit building my own computers many years ago. The components were cheap, and I always seemed to justify the amount of time it took to bring them all together into a stable, usable PC.

Slowly I began to understand the value of time. Time with friends and family. Time to watch my kids grow up. Time to build a business. Time for R&R.

I went from building my own boxes from scratch to modifying preconfigured boxes to simply buying a box and calling it a day. While I may no longer have the best of every single component inside the box what I have gained is a better quality of life and more time to focus on the things that are most important to me. And I can tell you what is most important is not sitting in the office tinkering with hardware. Unless it happens to be the short block from a 67 Mustang Fastback.

The real differentiators for the "cloud" itself, to me, will be ease of use. How quickly can I get up and running? How efficiently and cost effectively can I keep things running? And how reliably can I continue to meet my SLAs as I grow?

On top of that, can it support my favorite apps? And what sort of other apps of interest to me are available for the cloud of my choice?

Frankly, it's no different than the expectations of managed web hosting today. If it's easy to use, supports everything I need with the speed and reliability I want, at a price I can afford, I'm in.

[Wattersjames]

As someone who like Randy spends all of their time working on building clouds through service provider partners I'd like to share a quick quote that, I think, highlights some of Chuck's comments around Health-Care provision as proxy for some parts of the cloud market as well.

A large partner, with both a commodity Xen based cloud and a very large VMware hosting business said:

"Yes, we talk about Xen, all the time to developers, its hip its cool, its about taking on the man and the world. But when we get people with large budgets in to discuss their ears really never perk up till we mention VMware--then they get their wallets out. Because that's where they have placed their trust, and that's who they count to handle their business."

This is as literal of a quote as I can remember it.

[Chuck Hollis]

Insightful comment -- thanks for sharing!

-- Chuck

[Gavin Jackson]

In the argument of commodity vs differentiated and in the context of Service Providers and their business models, to my mind there is no argument... differentiated wins!

It is correct to say that Enterprises hold the cards and will vote, perhaps even tentitively, with their dollars/pounds/euros etc.

It would be a very brave Service Provider that leads with cost and commodity as it's core value.

Quite apart from the other considerations Enterprises have to make in terms of security, exit strategy, service interuption and subsequent service remedy (all relevent themes to support 'differentiated' clouds for SP's today), Service Providers will miss the next wave of revenue opportunity that is coming as a result of technology advances in the not too distant future.

When you look at the forthcoming products from Vmware (project Redwood for example) and the recent announcements by EMC for Global Active-Active cloud storage (V-Plex), the use-cases that become exposed for savvy Service Providers becomes game-changing!

We are of course talking about real 'federation'. Enterprises having the ability to toggle between internal and external services in a Private Cloud with consistent technologies and pooled resources.

I'm thinking Active (internal) - Active (external) environments for 'burst' resource allocation. I'm thinking follow the sun mobility for workloads and workers. I'm thinking M&A cloud integration services. I'm thinking global power calculators in real time helping Enterprises move from one location to another to benefit from power costs at any given time. There are several more, but you get the point.

These opportunities simply won't be open to commodity players. Enterprises won't have the ability to toggle between their Enterprise-grade infrastructure and the Service Providers bits and pieces. And frankly (IMHO) I don't think they'd want to!

In the final analysis, Service Provider business models that deal exclusively in commodities can only ever see cost lines going up due to sprawl of distributed infrastructure and revenue lines going down due to price-pressures. That's not a business model I'd be embracing in a hurry!

Gavin